Solid state disk life prediction method

Abstract

The invention discloses a solid state disk life prediction method. The method comprises the following steps of 1, collecting data to train a model; and 2, verifying, debugging and correcting the modelin the step 1 to obtain a trained model, and obtaining a threshold value. 3, inputting data in a hard disk to be predicted into the trained model in the step 2 to obtain a result; if the result is greater than the threshold value in the step 2, judging that the hard disk has a problem, giving an alarm, and updating the latest information of the hard disk; and if the result is smaller than the threshold value in the step 2, judging that the hard disk has no problem, and updating hard disk data. The method can more intelligently and accurately predict the hard disk service life according to theactual resource use condition of the user, is simple and flexible in algorithm, and can be quickly expanded to adapt to more resource use conditions.

Description

【Technical Field】 [0001] The invention belongs to the technical field of storage systems, and in particular relates to a method for predicting the life of a solid state hard disk. 【Background technique】 [0002] In large-scale distributed storage systems, there are many types of hard disks used, and the quality gap is large. In actual use, each disk is used in different ways. Therefore, it is easy to cause some disks to break and cause data loss. If given A hard disk life prediction technology, which can provide customers with a preventive technical guarantee. Existing solutions are based on the SMART (Self-Monitoring, Analysis and Reporting Technology) technology embedded in the hard disk device. The fault prediction accuracy rate provided is only about 30%, and no clear warning instructions can be given. The main reason for the failure of SMART prediction is that this technical framework started more than 20 years ago and is too simple. It is simply based on internal statistics...

AI Technical Summary

Problems solved by technology

[0002] In a large-scale distributed storage system, there are many types of hard disks, and the quality varies greatly. In the actual use process, each disk is used in a different way, so it is easy to cause some disks to be damaged and cause data loss. If given A hard disk life estimation technology, which can provide customers with a technical guarantee to prevent problems before they happen. The existing solutions are all based on the S.M.A.R.T (Self-Monitoring, Analysis and Reporting Technology) technology embedded in the hard disk device. The failure prediction accuracy rate provided is only about 30%, and it is impossible to give a clear early warning indication. The main reason for the failure of S.M.A.R.T prediction is that this technical framework started more than 20 years ago, which is too simple and is simply based on internal statistical data. Linear static analysis, in the practical application of ever-changing load patterns, does not have the ability to adjust and adapt at all, and the accuracy is greatly reduced. It is not enough to provide users with accurate prediction prompts, and may cause interference to users and even bring some economic losses.

Images